03-30-2018 10:24 AM
I have been a basic SAS user for a couple of years now. I have learned a lot from everyone in these communities. I am a certified Base Programmer, however, I use more PROC SQL than anything so I am very rusty on the SAS syntax. Most datasets I have created have been monthly snapshots of a loan portfolio over the past couple of years. They are a snapshot in time, not historical. I have these datasets from each month going backward.
I have been asked to start adding analysis to our reporting that tracks trends on things like delinquency as of each quarter for the past year and then project what we can expect over the next year. We use Tableau for these types of reports and I connect my SAS datasets to them and do the summarization within them. That being said, I don't really know where to start or where exactly I need to go to learn how to accomplish this type of reporting. A few questions that come to mind:
I primarily use SAS EG in my company so that is my primary tool. I have to learn how to do this somewhat quickly so I am hoping you guys can point me in the right direction to learn. I don't have a statistics background either, so that is another hurdle I must get over. I am just a lowly SQL programmer who has learned to use SAS for the past 2 years
As always, thanks for your help.
03-30-2018 12:46 PM - edited 03-30-2018 12:51 PM
I would let SAS do the heavy data lifting and use Tableau for presentation.
Look at this:
I have never used PROC ARIMA or PROC FORECAST but something like that would seem to be where you would want to go.
500K records a month? 6 million a year? SAS can probably do that actually.
Edit: Also this:
03-30-2018 01:00 PM
@HB, thanks for your reply. This looks like great information. I will take a look at that. My biggest concern with that many records is just how long it will take each month to run the reports with a dataset that large.
03-30-2018 01:05 PM
I run stuff on 1 million records all the time. The job I ran a few minutes ago looked like:
200 data orphans;
201 set mydata;
NOTE: There were 519500 observations read from the data set MYDATA.
NOTE: The data set WORK.ORPHANS has 519500 observations and 217 variables.
NOTE: DATA statement used (Total process time):
real time 10.96 seconds
cpu time 5.47 seconds
So 6 million records? Under a minute? I dunno.
03-30-2018 01:08 PM
Interesting. Do you submit jobs directly on the server? I actually run all of my queries from SAS Enterprise Guide on my desktop. I think that may cause a delay.
03-30-2018 01:23 PM
I don't use Enterprise Manager. I run 9.2. type a program, hit submit. It used to run faster but now we run it in a virtual desktop so it is just a bit slower.
03-30-2018 03:06 PM
Gotcha. Our company did away with SAS 9.2 and made us start using SAS EG full time. Just a different way to do the same thing. I will research some more and play with the large dataset to see how it performs.
03-30-2018 08:40 PM
Do you have access to a remote SAS server or is local SAS on your PC your only option? Once you start getting into millions of records then a remote SAS server is definitely the way to go. BTW the SAS product for forecasting is SAS/ETS.
04-06-2018 10:53 AM